Latent Spacecraft: Brains, GANs, Finnegans

This project argues that latent space—the learned, multidimensional, representational substrate of neural networks—is not only a mathematical construct but, read through literature, a structure akin to architectural interiority that renders the abstract physically legible. We stage three case studies to trace deep homologies between biological and synthetic language: (1) speech-producing brains alongside speech-generating GANs, which learn by informative imitation and exhibit imagitation, which exceeds mere replication; (2) Joyce’s Finnegans Wake as a cartography of latent space, where nonce words and associative syntax model pre-narrative interiority; and (3) FinneGANs, a GAN trained on Finnegans Wake audio, whose outputs probe the limits of comprehensibility and pre-speech form. Across these isomorphs, we contend that language is a final, linearizing layer over a generative, precategorial interior. Architecture serves as a navigable armature of the interior, where the symbolic layer is often physical, all the way to the deepest layers.